3D map building with mobile robots

Abstract

This paper presents a feature-based registration for 3D environments using mobile robots. The developed 3D laser scanner with custom hardware setup is able to scan both indoor and outdoor. For the map registration a nonlinear variant of the Iterative Closest Point (ICP) algorithm was used with initial alignment from the correspondences given by the features of the scenes. The initial alignment was determined using a set of key-points and the features of the keypoints in order to reduce the computational time and to ensure a robust estimation. Considering the increasing interest in 3D navigation for mobile robots, our aim was to use the created maps for both indoor and outdoor navigation purposes. Several maps were built by merging point clouds while our method was tested for a wide range of datasets including urban and office environments.